Radiographic Diagnostic Choices of Palestinian Dentists for Implant Therapy.

Aim: To report the radiographic preferences during dental implant therapy in Palestine. Materials and Methods: Fourteen multiple-choice questions were delivered in electronic and hardcopy formats questionnaires during the Sixth International Implantology Conference (Palestine). The questions investigated the radiographic techniques that are mostly used based on various clinical scenarios and treatment phases. Results: One hundred and thirty-seven responses were captured. The majority of the participants were general dentists with implant experience (79.6%). Less than a third of the participants (27.2%) were members of the Palestinian Association of Dental Implantology. The majority (85.9%) of them have their practice in a city zone. Panoramic radiograph (PAN) combined with cone beam computed tomography (CBCT) was the most preferred radiographic technique during the planning stage. Conclusion: PAN and CBCT was the preferred choice during the planning stages. A PAN was preferred postoperatively and if no complications were associated. In the case of symptomatic patients, CBCT was the radiograph of choice.

Dental age estimation: A comparative study of convolutional neural network and Demirjian's method.

The aim of this study is to compare a technique using Convolutional Neural Network (CNN) with the Demirjian's method for chronological age estimation of living individuals based on tooth age from panoramic radiographs. This research used 5898 panoramic X-ray images collected for diagnostic from pediatric patients aged 4-17 who sought treatment at Antalya Oral and Dental Health Hospital between 2015 and 2020. The Demirjian's method's grading was executed by researchers who possessed appropriate training and experience. In the CNN method, various CNN architectures including Alexnet, VGG16, ResNet152, DenseNet201, InceptionV3, Xception, NASNetLarge, InceptionResNetV2, and MobieNetV2 have been evaluated. Densenet201 exhibited the lowest MAE value of 0.73 years, emphasizing its superior accuracy in age estimation compared to other architectures. In most age categories, the predicted age closely matches the actual age. The most inconsistent results are observed at ages 12 and 13. The results highlight correspondence between the age predicted by CNN and the Demirjian's approach. In conclusion, the results show that the CNN method is adequate to be an alternative to the Demirjian's age estimation method. We suggest that convolutional neural network can effectively optimize the accuracy of age estimation and can be faster than traditional methods, eliminating the need for additional learning from experts.

Combining public datasets for automated tooth assessment in panoramic radiographs.

OBJECTIVE: Panoramic radiographs (PRs) provide a comprehensive view of the oral and maxillofacial region and are used routinely to assess dental and osseous pathologies. Artificial intelligence (AI) can be used to improve the diagnostic accuracy of PRs compared to bitewings and periapical radiographs. This study aimed to evaluate the advantages and challenges of using publicly available datasets in dental AI research, focusing on solving the novel task of predicting tooth segmentations, FDI numbers, and tooth diagnoses, simultaneously. MATERIALS AND METHODS: Datasets from the OdontoAI platform (tooth instance segmentations) and the DENTEX challenge (tooth bounding boxes with associated diagnoses) were combined to develop a two-stage AI model. The first stage implemented tooth instance segmentation with FDI numbering and extracted regions of interest around each tooth segmentation, whereafter the second stage implemented multi-label classification to detect dental caries, impacted teeth, and periapical lesions in PRs. The performance of the automated tooth segmentation algorithm was evaluated using a free-response receiver-operating-characteristics (FROC) curve and mean average precision (mAP) metrics. The diagnostic accuracy of detection and classification of dental pathology was evaluated with ROC curves and F1 and AUC metrics. RESULTS: The two-stage AI model achieved high accuracy in tooth segmentations with a FROC score of 0.988 and a mAP of 0.848. High accuracy was also achieved in the diagnostic classification of impacted teeth (F1 = 0.901, AUC = 0.996), whereas moderate accuracy was achieved in the diagnostic classification of deep caries (F1 = 0.683, AUC = 0.960), early caries (F1 = 0.662, AUC = 0.881), and periapical lesions (F1 = 0.603, AUC = 0.974). The model's performance correlated positively with the quality of annotations in the used public datasets. Selected samples from the DENTEX dataset revealed cases of missing (false-negative) and incorrect (false-positive) diagnoses, which negatively influenced the performance of the AI model. CONCLUSIONS: The use and pooling of public datasets in dental AI research can significantly accelerate the development of new AI models and enable fast exploration of novel tasks. However, standardized quality assurance is essential before using the datasets to ensure reliable outcomes and limit potential biases.

Evaluation of Styloid Process Elongation in Madinah, Saudi Arabia: A Retrospective Radiographic Investigation.

OBJECTIVES: This study aimed to identify the prevalence of an elongated styloid process and analyze the presence of its calcification in the Saudi population using panoramic radiographs. METHODS: The Taibah Outpatient Dental Clinic's OPG radiographs for 962 patients who attended screening clinics between December 2022 and October 2023 were all included in the study. Patients' demographics, such as age, gender, and nationality, as well as radiological data, were included in the following study variables: the presence of an elongated styloid on both sides of a panoramic radiograph, right side styloid length, left side styloid length, right side distal end thickness, and left side distal end thickness. RESULTS: The study evaluated 438 (45.5%) processes found in individuals aged 16-80 years old. The elongated process length ranged from 30.0 to 40.1 mm, and the diameter ranged from 0.81 to 7.79 mm at the origin to 0.56-3.79 mm at the end. There was no statistically significant difference in process length across genders or age groups. The diameters of the styloid bones on the left side vary significantly across genders at the start and completion of the process. CONCLUSION: The prevalence of elongated styloids in the studied population was 4.26%. The radiological evaluation of the styloid process is a crucial stage in dental surgery planning.

Panoramic imaging errors in machine learning model development: a systematic review.

OBJECTIVES: To investigate the management of imaging errors from panoramic radiography (PAN) datasets used in the development of machine learning (ML) models. METHODS: This systematic literature followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses and used three databases. Keywords were selected from relevant literature. ELIGIBILITY CRITERIA: PAN studies that used ML models and mentioned image quality concerns. RESULTS: Out of 400 articles, 41 papers satisfied the inclusion criteria. All the studies used ML models, with 35 papers using deep learning (DL) models. PAN quality assessment was approached in 3 ways: acknowledgement and acceptance of imaging errors in the ML model, removal of low-quality radiographs from the dataset before building the model, and application of image enhancement methods prior to model development. The criteria for determining PAN image quality varied widely across studies and were prone to bias. CONCLUSIONS: This study revealed significant inconsistencies in the management of PAN imaging errors in ML research. However, most studies agree that such errors are detrimental when building ML models. More research is needed to understand the impact of low-quality inputs on model performance. Prospective studies may streamline image quality assessment by leveraging DL models, which excel at pattern recognition tasks.

Prevalence and Distribution of Dental Anomalies in Schoolchildren in Kuwait.

OBJECTIVE: Dental anomalies (DA) can affect paediatric patients' aesthetics, function, and psychological well-being. There is a lack of data about the prevalence of DA in children in Kuwait. This study aimed to investigate the prevalence and distribution of DA amongst schoolchildren aged 8 to 12 years. METHODS: A retrospective study was conducted using panoramic digital radiographs of children who attended a single dental center. All radiographs were evaluated by 2 calibrated and trained examiners. RESULTS: DA were present in 110 (20.1%) out of the 546 panoramic radiographs examined: 53.6% in females and 46.4% in males. The mean age of children with DA (9.83 ± 1.29) was similar to that of children with no anomalies (9.96 ± 1.46). The most prevalent anomaly was dental agenesis (9.3%), followed by taurodontism (6.6%) and ectopic eruption (EE, 2%). DA were more common in the maxilla (58.2%) compared to the mandible (41.8%, P = .042). Congenitally missing teeth were significantly more frequent in the mandible (56.9%) than in the maxilla (43.1%, P = .003). EE was significantly more common in the maxilla (90.9%) than in the mandible (9.1%, P = .024). Microdontia and root dilacerations were only present in males, whilst supernumerary teeth, transposition, and impacted teeth were noted in females only. CONCLUSIONS: The prevalence of DA amongst schoolchildren in Kuwait was considered to be relatively high. Certain DA were associated with gender. The significant prevalence of DA highlights the need for early diagnosis using panoramic radiographs, particularly during the ages of 9 and 10, in order to ensure effective patient management.

Characteristics of CBCT versus panoramic signs of the surrounding bone of molar root protrusions into the maxillary sinus.

The study aimed to (a) investigate the amount and characteristics of the surrounding bone of protruded molar roots into the maxillary sinus using cone-beam computed tomography (CBCT) and (b) assess the correlation between the amount of bone with panoramic high-risk signs. Radiographs of 408 roots protruding beyond the sinus floor were evaluated. Axial CBCT images were used to investigate then classify eight characteristics of surrounding bone: no bone; bone < half the root girth in the proximal or buccal-palatal aspect; bone covering half the root girth in the proximal or buccal-palatal aspect; bone > half the root girth in the proximal or buccal-palatal aspect; and, complete bone. These were then grouped into four degrees of bone support: no bone; bone ≤ half the root girth; bone > half the root girth; and, complete bone. Panoramic signs were subclassified as: projection of root; interruption of the sinus floor; darkening of the root; upward curving of the sinus floor; absence of periodontal ligament space; and, absence of the lamina dura. Correlation between the degree of bone and the panoramic signs was evaluated using the Chi-square or Fisher's exact tests. Positive and negative predictive values, sensitivity, specificity, accuracy, and receiver operating characteristic analysis were calculated. Complete bone support was the most common. 'Projection of root' had a high negative predictive value and sensitivity. 'Absence of the periodontal ligament space and lamina dura' had a high positive predictive value, specificity, accuracy, and area under the curve. These two signs were significantly correlated with the degree of bone support.

Is there an association between molar incisor hypomineralization and developmental dental anomalies? A case-control study.

BACKGROUND: The aim of this study was to determine whether there is any association between molar incisor hypomineralization and developmental dental anomalies. METHODS: Two pediatric dentists evaluated panoramic radiographs of 429 children aged 8-14 years with molar incisor hypomineralization (study group) and 437 children without molar incisor hypomineralization (control group) in terms of developmental dental anomalies. Twelve different developmental dental anomalies were categorized into four types: size (microdontia, macrodontia); position (ectopic eruption of maxillary permanent first molars, infraocclusion of primary molars); shape (fusion, gemination, dilaceration, taurodontism, peg-shaped maxillary lateral incisors); and number (hypodontia, oligodontia, hyperdontia) anomalies. RESULTS: No significant difference was observed in the frequencies of developmental dental anomalies between the study and control groups in total, females, and males (p > 0.05). A statistically significant difference was found between the distribution of developmental size, position, shape, and number anomalies between the study and control groups (p = 0.024). The most common anomaly in both groups was hypodontia (6.3% and 5.9%, respectively). There was a significant difference between the study and control groups in terms of subtypes of shape anomaly in all children and females (p = 0.045 and p = 0.05, respectively). CONCLUSIONS: While a significant difference was observed between the distributions of types of developmental dental anomalies between individuals with and without molar incisor hypomineralization, there was no difference in terms of the frequency of developmental dental anomalies.

Accuracy of Anatomical Depictions in Cone Beam Computed Tomography (CBCT)-Reconstructed Panoramic Projections Compared to Conventional Panoramic Radiographs: A Clinical Risk-Benefit Analysis.

INTRODUCTION: Two-dimensional (2D) radiographs are the standard of care for diagnosis and treatment planning in the day-to-day practice of dentistry. With the growing popularity of cone beam computed tomography (CBCT), it is now becoming the standard of care in many areas of general dentistry due to its ability to create non-linear projections from volumetric data. The CBCT-generated non-orthogonal radiographs can serve as easy-to-use 2D and three-dimensional (3D) diagnostic tools and offer a similar experience for diagnosis as conventional 2D images. The aim of this study is to compare the accuracy of conventional radiographs and CBCT-generated projections to identify relevant anatomic landmarks and their associated variants. METHODS: Thirty-two patients referred to the University of Connecticut School of Dental Medicine's Advanced Imaging Center were selected for this retrospective analysis. Nineteen anatomical landmarks were retrospectively assessed on conventional panoramic and CBCT scans generated panoramic radiographs using two different digital imaging and communications in medicine viewers. A total of 1,216 anatomical landmarks were evaluated by two oral and maxillofacial radiologists to assess the accuracy and consistency of the depiction of radiographic anatomy. RESULTS: There was a very good agreement between the two evaluators with a Cohen's kappa value of 0.934. McNemar change test concluded that the anatomical assessment values compared between conventional panoramic and CBCT-generated panoramic radiographs are similar. CONCLUSION: This study showed that CBCT-generated panoramic images are comparable to conventional panoramic radiographs in identifying anatomical landmarks typically evaluated using a conventional panoramic projection. In addition, they have the added advantage of having 3D information in the acquired volume to better evaluate the area of interest. In clinical situations where a mid- to large-volume CBCT scan is available, a simulated panoramic image can be generated using the CBCT volume, leaving exposure of the patient to the additional radiation of a panoramic image unnecessary.

Deep learning and artificial intelligence in dental diagnostic imaging.

The application of artificial intelligence (AI) based on deep learning in dental diagnostic imaging is increasing. Several popular deep learning tasks have been applied to dental diagnostic images. Classification tasks are used to classify images with and without positive abnormal findings or to evaluate the progress of a lesion based on imaging findings. Region (object) detection and segmentation tasks have been used for tooth identification in panoramic radiographs. This technique is useful for automatically creating a patient's dental chart. Deep learning methods can also be used for detecting and evaluating anatomical structures of interest from images. Furthermore, generative AI based on natural language processing can automatically create written reports from the findings of diagnostic imaging.

Uso da radiografia panorâmica como fonte de informação para a identificação humana ­ relato de caso pericial / Use of panoramic radiograph as a source for human identification ­ forensic case report

Introdução: A odontologia legal permite a identificação humana por meio da comparação de dados observados em uma documentação odontológica ante mortem (AM) com as informações coletadas post mortem (PM), sendo os exames radiográficos grandes aliados neste processo. Objetivo: Demostrar a contribuição da radiografia panorâmica como fonte de informação para a identificação humana. Relato do caso: foi encaminhado um corpo carbonizado, com parte da região bucomaxilofacial preservada e que ao exame necroscópico era possível identificar a presença de restaurações e ausências dentais. A apresentação de radiografia panorâmica anterior à morte e a realização de exame radiográfico panorâmico no corpo carbonizado possibilitou a comparação de pontos coincidentes e divergências explicáveis, permitindo a identificação positiva do caso. Conclusão: Ao final da perícia foi determinada a identidade da vítima e foi comprovada a importância da radiografia panorâmica para a identificação humana com base em caracteres identificadores anatômicos e terapêuticos presentes no complexo bucomaxilofacial

Introduction: Forensic dentistry allows human identification through the comparison of data observed in ante-mortem (AM) dental documentation with information collected post-mortem (PM), with radiographic examinations being great allies in this process. Objective: To demonstrate the contribution of panoramic radiography as a source of information for human identification. Case report: a charred body was sent, with part of the oral and maxillofacial region preserved and upon necroscopic examination it was possible to identify the presence of restorations and missing teeth. The presentation of a panoramic radiograph prior to death and the performance of a panoramic radiographic examination of the charred body made it possible to compare coincident points and explainable divergences, allowing positive identification of the case. Conclusion: At the end of the forensic examination of the case, the identity of the victim was determined and the importance of panoramic radiography for human identification based on anatomical and therapeutic identifying characters present in the oral and maxillofacial complex was proven

Deep learning for sex determination: Analyzing over 200,000 panoramic radiographs.

The objective of this study is to assess the performance of an innovative AI-powered tool for sex determination using panoramic radiographs (PR) and to explore factors affecting the performance of the convolutional neural network (CNN). The study involved 207,946 panoramic dental X-rays and their corresponding reports from 15 clinical centers in São Paulo, Brazil. The PRs were acquired with four different devices, and 58% of the patients were female. Data preprocessing included anonymizing the exams, extracting pertinent information from the reports, such as sex, age, type of dentition, and number of missing teeth, and organizing the data into a PostgreSQL database. Two neural network architectures, a standard CNN and a ResNet, were utilized for sex classification, with both undergoing hyperparameter tuning and cross-validation to ensure optimal performance. The CNN model achieved 95.02% accuracy in sex estimation, with image resolution being a significant influencing factor. The ResNet model attained over 86% accuracy in subjects older than 6 years and over 96% in those over 16 years. The algorithm performed better on female images, and the area under the curve (AUC) exceeded 96% for most age groups, except the youngest. Accuracy values were also assessed for different dentition types (deciduous, mixed, and permanent) and missing teeth. This study demonstrates the effectiveness of an AI-driven tool for sex determination using PR and emphasizes the role of image resolution, age, and sex in determining the algorithm's performance.

Assessment of Odontometric and Osteologic Traits for Gender Determination in South Indians: An Original Research.

In forensic, odontologic, genetic, and anthropological aspects, odontometric and osteologic features have long been a valuable source. The goal of this research was to correlate both the osteologic and odontometric characteristics to determine the most accurate approach for determining gender. A retrospective study involving 1000 adults, with equal gender distribution, was carried out utilizing digital panoramic radiography. The archives were searched for radiographic images of the subjects that were procured for the various procedures that ranged from implantations to rehabilitations. The measurement process was carried out with Image-Pro. There was a noticeable gender difference in the mesodistal breadth, which ranged from 17 to 47. Asymmetry of the lower jaw was considerable in both genders, as was gender variance in osseologic characteristics including ramus diameter and gonial angle. The two groups of attributes exhibited a substantial positive predictive value and thus can be used indetermining gender.

Assessment of the correlation between mandibular radiomorphometric patterns and dual-energy X-ray absorptiometry measurements for screening bone mineral density.

The objective of this study was to investigate the correlation between scores for femoral and lumbar spine bone mineral density (dual-energy X-ray absorptiometry [DXA]) and visual, qualitative mandibular bone pattern assess¬ments (mandibular cortical index, trabecular bone pattern, and visual mandibular cortical width) as well as age and body mass index. Three trained observers evaluated 200 panoramic radiographs and 200 femoral and lumbar spine DXA measurements from 100 male and 100 female participants. The κ test showed an acceptable agreement among observers (0.73; P = 0.003). The Shapiro-Wilk test revealed that the variables were not normally distributed, so the Spearman correlation test was used. The mean age of the sample was 60.7 (13.9) years, and 86.0% of the patients were White. There were inverse correlations between the mandibular cortical index and the spine T-score in men, women, and the total sample as well as between the spine Z-score in the total sample. An inverse correlation was observed between the trabecular bone pattern and the spine T- and Z-scores in women and the total sample. The results also showed inverse correlations between the visual mandibular cortical width and all parameters analyzed in men, women, and the total sample except for the femur T-score and spine T- and Z-scores in men. Body mass index was correlated with all DXA parameters. Age was inversely correlated with femur T-scores in men and women but not with spine DXA values in men. The results suggest that qualitative assessments of radiomorphometric patterns on panoramic radiographs correlate with DXA values and therefore are suitable for screening patients at risk of developing low bone mineral density.

Assessment of Morphological Variations of the Coronoid Process, Condyle, and Sigmoid Notch as an Adjunct in Personal Identification Using Orthopantomograms Among the North Indian Population.

AIM: The aim of this study is to assess morphological variations of the coronoid process, condyle, and sigmoid notch as an adjunct in personal identification using orthopantomograms among the North Indian population. METHODOLOGY: The study sample (n=240) was distributed into four age groups: Group I: 30 males and 30 females aged 10-19 years, Group II: 30 males and 30 females aged 20-29 years, Group III: 30 males and 30 females aged 30-39 years, and Group IV: 30 males and 30 females aged 40-59 years. All were subjected to panoramic radiographs. The different morphological forms of the coronoid process, condyle, and sigmoid notch were evaluated. RESULTS: The results showed that across all age groups, angular condyles were the most common kind of condyle in males, followed by round and convex types. The present study found that the coronoid process typically takes on a triangle shape across all ages and sexes. Additionally, the vast majority of cases were triangular on both sides, and this was true across both sexes. It was found in this study that the sigmoid notch most commonly took the form of a larger notch, followed by a rounder notch. CONCLUSION: Using panoramic photos to portray the different morphologies of the coronoid process, condyle, and sigmoid notch can be a much simpler and faster method of identifying an individual, especially in the event of a mass disaster, so long as antemortem data are kept. The method of radiographic identification of individuals has recently gained prominence due to its efficacy. Radiographs like these can be invaluable in forensic dentistry, where they can help unearth previously hidden evidence if premortem records are retained. As a potential approach for individual identification among our population, panoramic radiographs were used to investigate the varying morphological forms of the coronoid process, condyle, and sigmoid notch.

Artificial Intelligence for Classifying the Relationship between Impacted Third Molar and Mandibular Canal on Panoramic Radiographs.

The purpose of this investigation was to evaluate the diagnostic performance of two convolutional neural networks (CNNs), namely ResNet-152 and VGG-19, in analyzing, on panoramic images, the rapport that exists between the lower third molar (MM3) and the mandibular canal (MC), and to compare this performance with that of an inexperienced observer (a sixth year dental student). Utilizing the k-fold cross-validation technique, 142 MM3 images, cropped from 83 panoramic images, were split into 80% as training and validation data and 20% as test data. They were subsequently labeled by an experienced radiologist as the gold standard. In order to compare the diagnostic capabilities of CNN algorithms and the inexperienced observer, the diagnostic accuracy, sensitivity, specificity, and positive predictive value (PPV) were determined. ResNet-152 achieved a mean sensitivity, specificity, PPV, and accuracy, of 84.09%, 94.11%, 92.11%, and 88.86%, respectively. VGG-19 achieved 71.82%, 93.33%, 92.26%, and 85.28% regarding the aforementioned characteristics. The dental student's diagnostic performance was respectively 69.60%, 53.00%, 64.85%, and 62.53%. This work demonstrated the potential use of deep CNN architecture for the identification and evaluation of the contact between MM3 and MC in panoramic pictures. In addition, CNNs could be a useful tool to assist inexperienced observers in more accurately identifying contact relationships between MM3 and MC on panoramic images.

Features of Dental Anomaly Patterns in Finnish children as seen in panoramic radiographs at the late mixed stage.

OBJECTIVE: To describe the prevalence of the developmental abnormalities involved in Dental Anomaly Patterns (DAP) and investigate their co-occurrence in an age cohort of children with late mixed dentition. MATERIAL AND METHODS: Retrospective, register-based study focused on 1315 panoramic radiographs of children aged 8.5-10.5 years. The features examined were absent teeth, peg-shaped maxillary lateral incisor, delayed dental age, infraocclusion of primary molars, transposition and distal angulation of unerupted mandibular second premolar. RESULTS: Feature involved in DAP was detected in 29.8% of the children, most common being infraocclusion of primary molars (17.5%), followed by absent teeth (8.4%), delayed dental age (7.6%), distal angulation of unerupted mandibular second premolar (7.3%), peg-shaped maxillary lateral incisor (2.4%) and transposition (0.5%). Two DAP features occurred together in 4.7% of children, while three occurred in 0.7%. Infraocclusion (p=.040) and absent teeth (p=.001) occurred more commonly in girls. Phenotypic variations in maxillary lateral incisor more often occurred together (p=.004). Absent teeth, peg-shaped maxillary lateral incisor and delayed dental age more often occurred together (p<.01) as did transposition and absent teeth (p=.016). CONCLUSION: Almost third of the children had dental developmental abnormalities involved in DAP. Absent teeth, peg-shaped lateral incisors and delayed dental age more often occurred together.

Robust automated teeth identification from dental radiographs using deep learning.

OBJECTIVES: This study developed and validated a deep learning-based method to automatically segment and number teeth in panoramic radiographs across primary, mixed, and permanent dentitions. METHODS: A total of 6,046 panoramic radiographs were collected and annotated. The dataset encompassed primary, mixed and permanent dentitions and dental abnormalities such as tooth number anomalies, dental diseases, dental prostheses, and orthodontic appliances. A deep learning-based algorithm consisting of a U-Net-based region of interest extraction model, a Hybrid Task Cascade-based teeth segmentation and numbering model, and a post-processing procedure was trained on 4,232 images, validated on 605 images, and tested on 1,209 images. Precision, recall and Intersection-over-Union (IoU) were used to evaluate its performance. RESULTS: The deep learning-based teeth identification algorithm achieved good performance on panoramic radiographs, with precision and recall for teeth segmentation and numbering exceeding 97%, and the IoU between predictions and ground truths reaching 92%. It generalized well across all three dentition stages and complex real-world cases. CONCLUSIONS: By utilizing a two-stage training framework with a large-scale heterogeneous dataset, the automatic teeth identification algorithm achieved a performance level comparable to that of dental experts. CLINICAL SIGNIFICANCE: Deep learning can be leveraged to aid clinical interpretation of panoramic radiographs across primary, mixed, and permanent dentitions, even in the presence of real-world complexities. This robust teeth identification algorithm could contribute to the future development of more advanced, diagnosis- or treatment-oriented dental automation systems.

Artificial intelligence in the diagnosis of dental diseases on panoramic radiographs: a preliminary study.

BACKGROUND: Artificial intelligence (AI) has been introduced to interpret the panoramic radiographs (PRs). The aim of this study was to develop an AI framework to diagnose multiple dental diseases on PRs, and to initially evaluate its performance. METHODS: The AI framework was developed based on 2 deep convolutional neural networks (CNNs), BDU-Net and nnU-Net. 1996 PRs were used for training. Diagnostic evaluation was performed on a separate evaluation dataset including 282 PRs. Sensitivity, specificity, Youden's index, the area under the curve (AUC), and diagnostic time were calculated. Dentists with 3 different levels of seniority (H: high, M: medium, L: low) diagnosed the same evaluation dataset independently. Mann-Whitney U test and Delong test were conducted for statistical analysis (ɑ=0.05). RESULTS: Sensitivity, specificity, and Youden's index of the framework for diagnosing 5 diseases were 0.964, 0.996, 0.960 (impacted teeth), 0.953, 0.998, 0.951 (full crowns), 0.871, 0.999, 0.870 (residual roots), 0.885, 0.994, 0.879 (missing teeth), and 0.554, 0.990, 0.544 (caries), respectively. AUC of the framework for the diseases were 0.980 (95%CI: 0.976-0.983, impacted teeth), 0.975 (95%CI: 0.972-0.978, full crowns), and 0.935 (95%CI: 0.929-0.940, residual roots), 0.939 (95%CI: 0.934-0.944, missing teeth), and 0.772 (95%CI: 0.764-0.781, caries), respectively. AUC of the AI framework was comparable to that of all dentists in diagnosing residual roots (p > 0.05), and its AUC values were similar to (p > 0.05) or better than (p < 0.05) that of M-level dentists for diagnosing 5 diseases. But AUC of the framework was statistically lower than some of H-level dentists for diagnosing impacted teeth, missing teeth, and caries (p < 0.05). The mean diagnostic time of the framework was significantly shorter than that of all dentists (p < 0.001). CONCLUSIONS: The AI framework based on BDU-Net and nnU-Net demonstrated high specificity on diagnosing impacted teeth, full crowns, missing teeth, residual roots, and caries with high efficiency. The clinical feasibility of AI framework was preliminary verified since its performance was similar to or even better than the dentists with 3-10 years of experience. However, the AI framework for caries diagnosis should be improved.

[Application of Deep Learning in Differential Diagnosis of Ameloblastoma and Odontogenic Keratocyst Based on Panoramic Radiographs].

Objective To evaluate the accuracy of different convolutional neural networks (CNN),representative deep learning models,in the differential diagnosis of ameloblastoma and odontogenic keratocyst,and subsequently compare the diagnosis results between models and oral radiologists. Methods A total of 1000 digital panoramic radiographs were retrospectively collected from the patients with ameloblastoma (500 radiographs) or odontogenic keratocyst (500 radiographs) in the Department of Oral and Maxillofacial Radiology,Peking University School of Stomatology.Eight CNN including ResNet (18,50,101),VGG (16,19),and EfficientNet (b1,b3,b5) were selected to distinguish ameloblastoma from odontogenic keratocyst.Transfer learning was employed to train 800 panoramic radiographs in the training set through 5-fold cross validation,and 200 panoramic radiographs in the test set were used for differential diagnosis.Chi square test was performed for comparing the performance among different CNN.Furthermore,7 oral radiologists (including 2 seniors and 5 juniors) made a diagnosis on the 200 panoramic radiographs in the test set,and the diagnosis results were compared between CNN and oral radiologists. Results The eight neural network models showed the diagnostic accuracy ranging from 82.50% to 87.50%,of which EfficientNet b1 had the highest accuracy of 87.50%.There was no significant difference in the diagnostic accuracy among the CNN models (P=0.998,P=0.905).The average diagnostic accuracy of oral radiologists was (70.30±5.48)%,and there was no statistical difference in the accuracy between senior and junior oral radiologists (P=0.883).The diagnostic accuracy of CNN models was higher than that of oral radiologists (P<0.001). Conclusion Deep learning CNN can realize accurate differential diagnosis between ameloblastoma and odontogenic keratocyst with panoramic radiographs,with higher diagnostic accuracy than oral radiologists.